Détermination d'indices de stabilité et validation en classification non supervisée

Abstract

One important difficulty in automatic clustering is the validation of the results given by an algorithm. \\ In this report we will develop some indexes to estimate the stability of clustering results. Furthermore, we will investigate the use of these validation indexes in the determination of the "optimal" number of clusters.

Date of Award	26 Jun 2007
Original language	French
Supervisor	Andre Hardy (Supervisor), Jean Paul Rasson (Jury) & Marcel Remon (Jury)